Package installation incredibly slow unless in rescue mode

Latest response

Installing packages via yum (and other tools) is incredibly slow, even if I am installing from a repository on the local network. Yum install, yum erase, and yum reinstall all shows this problem. DNF has the same problem. I am not sure if DNF and Yum call the rpm command in the background (does not appear to be the case.) "Rpm --erase" is also very slow.

The system monitor does not show hi CPU, disk or network utilization. File copies to/from the network with scp are quite fast.

If I boot from DVD into rescue mode, then "chroot /mnt/sysimage" and "ifup em1" then the further installs using yum are quite fast.

Responses

I am also encountering this problem on new 48 cores system with 384GB ram, 2TB raid0 SSD volume. Installing 20mb package like kernel-debug-devel took 5 minutes minutes.

I've tested RHEL 7.9, RHEL 8.4 and CentOS 7.9. System does not exhibit this problem when running CentOS 7. I have never encountered this problem with CentOS.

Hello to both of you, Damian Lock and Pedro Davila.

Damian Lock, - it may be difficult to get to the bottom of the issue you present because there's quite a lot of possibilities. You mention - We do not know what kind of network you have, if it's pure virtual, if it's a tiny network, or if it is on a business network. My initial thought is to examine your /etc/resolv.conf - and I'm not sure if you do or do not have a local DNS server. When I see this sort of issue, I immediately doubt the correct contents of /etc/resolv.conf. -- I'm also curious if you do or do not have some network contention perhaps, that would take some work to get to the bottom. I'm curious also if other systems are affected by this, or if you only have one "client" system, namely the one you mention. I'm curious if the system you're attempting to get rpms from is being "dogpiled" and is under a heavy load perhaps. I'm curious if you were to do a "yumdownloader tmux" and attempt to install it locally, if that would work, and/or if (as a temporary test) make a local repository on the system itself would prove different (and if you do this, disable the repo file in /etc/yum.repos.d/ that points to the other yum repository. One thing is for certain, there is an obvious difference from when your system is running not in rescue mode, and also in rescue mode. Given the possible amount of things involved, we may recommend opening a case with Red Hat and provide them an sosreport.

Maybe for both of you - consider an entry in /etc/hosts with the ip/hostname of the system you are getting updates from, however in Pedro's case, we have no idea if he's using Red Hat's repos, or CentOS, or something nearby.

Pedro Davila, Some of the similar things I mentioned above for Damian Lock might apply to you, but you do not mention in your specific matter if you are using CentOS (I almost believe you are using either RHEL 7 or 8 however), and if you are or are not using a nearby "local" (but not-on-that-system) repository, or if you are going across the Internet to Red Hat. This makes a significant difference and details make a difference for which version of RHEL, if you are using a local-nearby-not-on-the-local-system-repo, or a on-the-system-repo, or repos at Red Hat all of this makes a difference.

It may go easier for us to help both of you to have separate discussions (Pedro, recommend starting a new discussion), because having one discussion where potentially others dogpile their "me too" things, can make a not-easy-to-read discussion.

Kind Regards,
RJ